movement operations after syntax - linguisticsembick/move.pdfsyntax david embick rolfnoyer we...

Movement Operations afterSyntaxDavid EmbickRolf Noyer

We develop a theory of movement operations that occur after thesyntactic derivation in the PF component within the framework ofDistributedMorphologyThe theory is an extensionof what was calledMorphological Merger in Marantz 1984 and subsequent work A pri-mary result is that the locality properties of a Merger operation aredetermined by the stage in the derivation at which the operation takesplace specifically Merger that takes place before Vocabulary Inser-tion on hierarchical structures differs from Merger that takes placepostndashVocabulary Insertionlinearization Specific predictions of themodel are tested in numerous case studies Analyses showing the inter-action of syntactic movement PF movement and rescue operationsare provided as well including a treatment of English do-support

Keywords morphologysyntax MorphologicalMerger adjacencyPFmovement Distributed Morphology

The properties of syntactic movement have been studied extensively in linguistic theory both interms of locality conditions and in terms of the types of constituents affected (phrases subpartsof phrases heads) Despite differences in particular analyses or frameworks the locality conditionson movement operations are a central concern of current research Here we address movementoperations as well but operations of a different type In particular we examine and analyzemovement operations that occur after the syntactic derivation in the PF component and providea theory that makes proposals concerning (a) the locality conditions on such movements (b) thetypes of constituents they affect and (c) the position of such operations in the sequential derivationfrom the output of syntax to phonologically instantiated expressions

From a somewhat abstract perspective the fact on which we base our study is that not allstructures and strings are the result of operations that occur exclusively in the syntactic componentof the grammar this observation stems from a body of prior research investigating the relationshipbetween syntactic structure and phonological form The observation covers two domains onedealing with linear sequences that are syntactically opaque the other with movement operationsIn the first domain it has been demonstrated that the internal ordering of clitic clusters cannot

We would like to thank Artemis Alexiadou Karlos Arregi Rajesh Bhatt Jonathan Bobaljik Morris Halle HeidiHarley Tony Kroch Alec Marantz David Pesetsky Beatrice Santorini and audiences at NYU MIT and ZASBerlinwhere parts of this article were presented We would also like to acknowledge the Penn Research Foundation for financialsupport for our research

555

Linguistic Inquiry Volume 32 Number 4 Fall 2001555ndash595

q 2001 by the Massachusetts Institute of Technology

556 D A V I D E M B I C K A N D R O L F N O Y E R

follow from syntactic principles (see Perlmutter 1971 also see Bonet 1991 and Noyer to appearfor explicit discussion in the framework assumed here) In the second domain which we explorehere it has also been recognized that syntactic movement cannot be responsible for certain move-ment operations One of the most familiar comes from the inflectional system of English Whileit has long been recognized that English main verbs do not move to T(ense) in the syntacticderivation (Emonds 1978 Pollock 1989) it is nevertheless the case that tense morphology appearson the verb in the surface string John kick-ed the ball to Mary The positioning of tense morphol-ogy on the verb has been attributed to Lowering a movement operation that ultimately derivesfrom the Affix Hopping transformation of early generative grammar Subsequent attempts toidentify and analyze lsquolsquomismatchesrsquorsquo between morphology and syntax can be found in the workof Marantz (1984 1988) and Sadock (1991) Our goal in this article is to develop a more compre-hensive theory of the types of morphological movement operations in the grammar and to situatethese operations within an explicit architecture for the PF derivation The focus is on cases inwhich some sort of displacement must apply to generate the proper phonological form from anidentifiable syntactic structure We thus exclude from consideration cases in which for examplethe syntactic distribution of clitic clusters is exactly what one would expect given the propersyntactic analysis1

1 Syntactic versus Postsyntactic

Our initial discussion of movement in PF has two components The first concerns why postsyntac-tic displacements are needed in addition to syntactic movement The second addresses the ideathat syntactic movements of the more familiar sort are part of the PF derivation as well

11 Clitic Placement

The need for nonsyntactic movements has been clearest in the domain of clitic placementmdashinparticular in the analysis of second-position clitics In some cases the question that was posedwas whether or not syntactic movement could possibly account for patterns of clitic placementor in particular second-position phenomena We take this question to be ill formed Syntax is agenerative system and assuming that one is willing to loosen many of the constraints on syntacticmovement the ability of such a system to capture certain linear orders should never have beenin question Rather the question that we take to be central is whether or not it is desirable tohave syntax perform such operations The two positions that we wish to contrast along with themeans by which each approach captures apparent postsyntactic movements are as follows

(1) a Syntax only Syntax performs operations that are explicitly executed so as to resolvea morphophonological problem Patterns of apparent postsyntactic movement arereducible to the effects of these lsquolsquospecialrsquorsquo syntactic processes

b PF movement Syntax generates and moves terminals according to its own principles

1 A third type of difference in structure between PF and syntax proper involves the addition of morphemes at PFto meet language-particular requirements this will be outlined below

M O V E M E N T O P E R A T I O N S A F T E R S Y N T A X 557

and is oblivious to morphophonologicalconcerns PF takes the output of syntax andresolves morphophonological dependencies according to its own principles

The position we adopt endorses the latter possibility2 In essence the idea in the domain ofclitic placement is that unless syntax incidentally provides a host for a clitic PF can performmovement operations to satisfy a clitic dependency (see eg Marantz 1984 1988 Halpern 1992bSchutze 1994 Embick and Izvorski 1995) Although the discussion of this section might seemsomewhat peremptory the literature on cliticization is vast and cannot be addressed here Argu-ments against syntactic treatments of various clitic placement phenomena are abundant in theliterature In the body of the article we will not make a lsquolsquonot syntacticrsquorsquo argument for each casestudy we present although such arguments could certainly be derived from the examples weanalyze Rather than focusing on this aspect of the phenomena we will provide a theory of thelocality conditions on postsyntactic readjustments of the type outlined above

12 Syntax in PF

Recent developments in syntactic theory particularly those associated with the Minimalist Pro-gram attempt to circumscribe the operations that syntax proper is supposed to perform Althoughwe do not adopt some of the more extreme versions of this view such as Chomskyrsquos (2001)position that head movement does not take place in syntax we find the general picture provided bysuch theoretical contexts compelling particularly to the extent that they acknowledge movementoperations on the PF branch This having been said the nature of the movements that we intendto discuss must be clarified somewhat Chomsky (2001) takes the more extreme position that agreat deal of apparently phrasal movement also takes place in the phonological component (insome cases this is what has been referred to as lsquolsquostylistic movementrsquorsquo in other cases not) Unlikethe movement operations to be studied here which have an immediately local character and whichare motivated by the satisfaction of primarily morphologicalor morphophonologicalrequirementsthe movements Chomsky relegates to PF seem to have many of the properties of straightforwardsyntactic movement

We take the position that the grammar includes only one syntactic component that is every-thing that looks like syntactic composition or movement takes place in syntax movement of thistype is not distributed across PF as well In part this position is motivated by parsimony twomodularly distinct syntactic systems should not be posited unless absolutely necessary Until ithas been conclusively demonstrated that a syntaxlike movement system is required at PF we willassume the more restrictive option

2 Background Assumptions

In this section we sketch our background assumptions about the structure of the grammar Weassume a theory in which morphology interprets the output of the syntactic derivation (Distributed

2 Much of the discussion of these issues centers around second-position cliticization effects especially in the Slaviclanguages Franks and Holloway King 2000 contains many relevant references to the literature


Morphology Halle 1990 Halle and Marantz 1993 1994 Noyer 1997 and related work) Thebasic elements of the syntactic computation are abstract features which appear in bundles asterminal nodes Universal as well as language-specific well-formedness conditions determine inwhat manner these features may combine to form syntactic categories the atoms of syntacticrepresentation which we call here morphemes We reserve the term morpheme in this model todenote terminal elements in syntactic or morphosyntactic representations we use the term expo-nent to denote the phonological expression(s) of a morpheme Overt syntax denotes those opera-tions that assemble and manipulate this collection of morphemes into a hierarchical structure thatis the input to computations concerned with morphological and phonological expression (PF) andaspects of semantic interpretation (LF)

The kinds of operations that take a representation from the PFLF branching to its phonologi-cal form comprise the (PF) Spell-Out of that representation the module of grammar responsiblefor Spell-Out is Morphology That is we use Morphology here as a cover term for a series ofoperations that occur on the PF branch following the point at which the syntactic derivation splitsbetween PF and LF There is no Lexicon in this model of grammar word formation takes placeeither in syntax or through postsyntactic operations during Morphology

21 Morphemes and Exponents

A number of independent operations occur on the PF branch and some of these are of directrelevance to the present study To begin with not all morphemes relevant to pronunciation arepresent in syntax prior to Spell-Out and Morphology In other words not all word constituentsare syntactic entities many are not and are purely morphological (Harris 1991 Noyer 1997)Certain structural positions within words and perhaps phrases are inserted in Morphology subjectto various conditions For example many languages require noun and adjective stems to beaugmented by case morphemes that are not themselves syntactic projections These morphemesmust be added postsyntactically during Morphology as in (2)

(2) Noun [Noun ` CaseN o u n ]

In the terminology of Embick 1997 such inserted morphemes are called dissociated sincethe information their signalization conveys is partly separated from the original locus of thatinformation in the phrase marker Typical dissociated morphemes include case and agreementmorphemes in at least some languages (see eg Marantz 1992) these morphemes reflect certainsyntactic properties (or configurations) but do not in any sense contribute these properties tosyntax Dissociated morphemes are not interpreted at LF since they are inserted only at Spell-Out

Following Halle and Marantz (1993) we adopt the term Vocabulary to denote the list ofphonologicalexponentsand their privileges of occurrence Each Vocabulary item is a phonologicalrepresentation paired with a set of conditions on its insertion Vocabulary items are inserted intoterminal nodes following the syntactic derivation because terminals are only provided with spe-cific Vocabulary items postsyntactically this is referred to as Late Insertion


We use the term morphemes or equivalently heads to refer to the basic atoms of phrasestructure which consist of abstract morphemesndashthat is bundles of features A variety of opera-tions includinghead adjunctionin syntax morpheme insertion in Morphologyand the postsyntac-tic Mergers to be studied below can produce X0 elements with internal structure such as in (3)

(3) Z adjoined to Y Z Y adjoined to X

XP

X WP

Y

Z Y

X

gd

ab

1

When Vocabulary Insertion occurs the Vocabulary is searched for items to be inserted intothe structure For certain morphemes Vocabulary Insertion is deterministic and only one choiceis possible in any given context Following Harley and Noyer (1998) we will call such morphemesf-morphemes typically such morphemes correspond to lsquolsquofunctionalrsquorsquo projections in syntax Forf-morphemes the Vocabulary item with the largest subset of the features present on the terminalnode is inserted3 For other morphemes there exists a choice as to which may be inserted wecall such morphemes Roots symbolized with the notation Iuml ROOT (see Pesetsky 1995) Unlikethe exponents inserted into f-morphemes the Roots are not in competition with each other andany Root licensed in a given environment may be inserted4

More concretely suppose that in (3) Z 4 Noun Y 4 Numberdual and X 4 Caseinstrumen-tal5 In the Vocabulary of Mansi a Uralic language the most specific Vocabulary items forNumberdual and Caseinstrumental will be aring and tl respectively (Kalman 1965) For the Rootposition there exist many choices naturally but the following forms exemplify possible outcomes

(4) a put-aring-tlkettle-DUAL-INST

lsquoby means of two kettlesrsquo

3 See Halle 1997 but certain complexities arise in defining the precedence relations among competing Vocabularyitems (see Harley and Noyer 19995 for discussion)

4 For different perspectives on the Late Insertion of Roots see Halle 1990 Marantz 1995 Harley and Noyer 1999and Embick 2000

5 Whether or not the hierarchical structure presented in (3) is the one that is needed for this case depends uponfurther assumptions about Number and Case which need not be addressed at this point


b eraring-aring-tlsong-DUAL-INST

lsquoby means of two songsrsquo

The architecture of the theory establishes a direct connection between syntax and morphol-ogy In the default case the structure interpreted in Morphology will simply be the output of thesyntactic derivation As is well known however the PF branch has its own properties and simplyas a matter of fact there are sometimes apparent mismatches between morphological and syntacticstructures Distributed Morphology accounts for such mismatches in terms of specific operationsthat occur in the PF derivation For instance the operation of Fission (Noyer 1992 Halle 1997)produces two terminal nodes from a single terminal node The status of such operations and theconditions on their application is the subject of an ongoing research program The project of thisarticle is to focus on one specific type of morphologysyntax mismatch involving PF movementit is important to emphasize however that the present research is situated in a theoretical frame-work in which morphological structure is unless further operations apply at PF simply syntacticstructure

22 Dependent Elements

The breakdown of morphemes exponents and operations sketched above provides the basicinventory for the theory we will develop here However further comments are in order concerninghow the ontology of our approach relates to distinctions found elsewhere in the literature Theelements involved in the movement operations that we examine are often classified as clitics oras affixes depending upon certain criteria Most salient is the distinction proposed in Zwicky andPullum 1983 and related work in which a number of properties are proposed in order to definethese two types of objects According to a line of reasoning prevalent in Lexicalist theoreticalframeworks clitics must be placed by syntax whereas affixes are placed on their hosts via Lexicalrules In the context of the present theory the modular distinction between clitics and affixescannot be maintained That is there is no Lexical mode of derivation in the present theory sosuch a distinction loses much of its importance Instead elements that appear bound to others onthe surface have heterogeneous derivational histories with clitic and affix being descriptive termsfor some of these The important factor in any particular analysis is therefore not whether anelement is a clitic or an affix Rather one must identify (a) the provenance of the morpheme(syntactic or dissociated) and its distribution and correspondingly (b) the means by which itcomes to be attached to its host Although we assume a non-Lexicalist model and believe theincorrectness of Lexicalism to have been amply demonstrated in the literature we will haveoccasion to illustrate the inadequacy of the cliticaffix 4 syntacticLexical equation in our casestudy of the Bulgarian DP below

23 Morphological Merger

MorphologicalMerger as first proposed by Marantz was originallya principle of well-formednessbetween levels of representation in syntax in Marantz 1988261 Merger is generalized as follows


(5) Morphological MergerAt any level of syntactic analysis (D-Structure S-Structure phonological structure) arelation between X and Y may be replaced by (expressed by) the affixation of thelexical head of X to the lexical head of Y

In this formulation Merger essentially lsquolsquotradesrsquorsquo or lsquolsquoexchangesrsquorsquo a structural relation betweentwo elements at one level of representation for a different structural relation between the lsquolsquoheadsrsquorsquoof these elements at a subsequent level Exactly what relations may be traded is on this conceptiondependent upon the levels of the grammar related by the operation In one case the relevant notionof headedness in terms of which the operation applies is defined linearly in another in terms ofstructure

The theory that we present here follows the insight that Merger is an operation with differentdomains of applicationThis view is situated in a theory in which Mergers taking place at differentstages in a sequential PF derivation have different locality properties The proposal is that thereare in fact at least two varieties of Merger depending upon whether Merger occurs (a) in Morphol-ogy before Vocabulary Insertion or (b) in Morphology after or concomitant with VocabularyInsertion6 The Merger of type (a) is Lowering it operates in terms of hierarchical structure TheMerger of type (b) Local Dislocation operates in terms of linear adjacency

3 Lowering versus Local Dislocation

We turn now to initial definitions and illustrations of our movement operations For expositorypurposes we will include a marker t to indicate the position of origin of an element affected byMerger This is not intended to represent a trace (or a copy) in the technical sense in which tracesbehave in syntactic theory Rather its purpose is merely to illustrate the nature of our movementoperations which occur after syntax and therefore do not leave traces or their equivalents

31 Lowering

Because we adopt the view that phonological expression of complex words is determined byinformation provided by the syntactic derivation in certain instances Lowering movement willbe required to unite syntactic terminals that are phonologically spelled together but not joined inovert syntax (by Raising) Here the head X lowers to Y the head of its complement7

(6) Lowering of X0 to Y0

[X P X0 [Y P Y0 ]] [X P [Y P [Y 0 Y0`X0 ] ]]

In English as opposed to a number of other languages V does not move to T in overt syntax

6 Another potential type of Merger Prosodic Inversion (Halpern 1992b) involves the trading of relations at thelevel of prosodic domains such as Phonological Word and Phonological Phrase We will not be discussing ProsodicInversion in detail here but it is easily incorporated within our proposals as a variety of Merger operating immediatelybefore Phonology

7 A possible refinement to this definition of Lowering will be suggested in section 72


However T is realized on V morphologically (except when negation appears or when T-to-Cmovement has occurred) English T thus lowers to V the head of its complement8

(7) a Mary [T P t1 [v P loudly play-ed1 the trumpet]]b Mary did loudly play the trumpet

Because Lowering involves adjunctionof a head to a head and these heads need not necessar-ily be linearly adjacent Lowering has a (potentially) nonlocal that is nonadjacent character AsBobaljik (1994) discusses an intervening adjoined adverb such as loudly in (7a) does not preventT from lowering to V In sum this postsyntactic operation is one that skips potentially interveningadjuncts and adjoins T to the head of its complement (ie v)

32 Local Dislocation

A second variety of Merger which we term Local Dislocation occurs after Vocabulary InsertionIn Local Dislocation the relation relevant for lsquolsquoaffixationrsquorsquo is not hierarchical but rather linearprecedence and adjacency By hypothesis linear ordering is not a property of syntactic representa-tions but is imposed at PF in virtue of the requirement that speech be instantiated in time (seeSproat 1985) It is therefore natural to assume that linear ordering is imposed on a phrase markerat the point in the derivation when phonological information is inserted that is at VocabularyInsertion

(8) The Late Linearization HypothesisThe elements of a phrase marker are linearized at Vocabulary Insertion

Our idea that Local Dislocation is a variety of Merger distinct from Lowering is based onMarantzrsquos (1988) position that the notion lsquolsquohead of a constituentrsquorsquo relevant to Merger is defineddifferently at different levels of the grammar The implementation of this position here is that theproperties of Merger differ depending upon whether Merger applies on a linearized or unlinearizedstructure Specifically before linear order is imposed on a phrase marker headedness is definedin terms of structure where a constituent C 4 X(P) then the head of C is X0 After linearizationthis no longer holds and the head is defined in terms of peripherality within the constituent Tosee how this is so consider the following hierarchical structure

(9) [X P X [Y P [Z P Z] Y]]

Here X takes YP 4 [ZP Y] as its complement where ZP is either a complement to Y or anadjunct to YP We will use the notation a b to denote a requirement that a must linearly precedeb and be adjacent to b A potential linearization of this structure is this

(10) [X [Z Y]]

8 On a strict Lexicalist theory such as that assumed in Chomsky 1995 play-ed is fully inflected before syntax inthe Lexicon its tense features need not be checked until LF When Spell-Out occurs play-ed already has the inflectionalmaterial with it However Distributed Morphology admits no Lexicon in which play-ed might be constructed beforesyntax


Here X must immediately precede [Z Y] and Z must immediately precede Y In the syntacticstructure (9) from which (10) originates Y is the lsquolsquoheadrsquorsquo of the constituent that X takes as itscomplement In syntax Y could raise to X likewise X could lower to Y across ZP But LocalDislocation does not refer to (9) rather it refers to (10) The relationship that is relevant is nowone of linear precedence and adjacency ( ) Specifically Local Dislocation can convert (10) to(11)

(11) [[Z 0 Z`X] Y]

In (11) Xrsquos relation to [Z Y] has been exchanged for a relation of adjunction to the left-peripheral element of [Z Y] namely Z (11) is a legitimate transformation of (10) because both relations in (10) have been either respected or properly converted by Local Dislocation9 In(10) Y must immediately follow and be adjacent to Z In (11) this relationship is maintainedbecause Y still follows Z0 which is now however internally complex as Z`X

As Marantz (1988) shows the idea that Local Dislocation Merger exchanges relations ofadjacency for those of adjunction places restrictions on the structures in which two elements canbe inverted in the string Our interpretation of these restrictions is as follows if X is an elementperipheral in some constituent C X will not be able to invert with an element Y that is outsideof the constituent C (12b) although leaning is possible (12c) Consider the following cases

(12) a [ Y] [C X Z]b [ X`Y] [C Z] impossible inversionc [ Y`X] [C Z] possible leaning

Given a prendashLocal Dislocation structure such as (12a) X cannot lsquolsquoescapersquorsquo the constituent C toinvert with Y as in (12b) since in so doing X would not properly maintain its requirement of(left-)adjacency with Z This should be contrasted with the derivation of (11) where X invertswith an element (Z) that is contained within the constituent that X is originally peripheral in

As we will illustrate throughout the article the domain in which these Mergers apply is notlimited to absolute sentence-peripheral position as originally envisioned by Marantz RatherMergers apply within particular domains (eg DPs) such that peripheral elements within suchdomains undergo Merger with other elements in that domain In such cases it appears that materialoutside the domain is irrelevant In (13) for example X may undergo Merger with Y such thatthe (cross-domain) relationship Z X is violated (domain boundary illustrated with | for clarity)

(13) Z | [X P X Y W] Z | [X P Y`X W]

The same type of effect is found with Lowering Thus for instance the head of DP maylower to its complement as we show below without incurring any violation from having removedits relationship with a dominating verb Whether or not these domains correspond to (or shouldcorrespond to) syntactically motivated subunits of structure such as the phases introduced inChomsky 2000 2001 is an open question of substantial interest

9 The affixation of X to Y might first involve rebracketing under adjacency such that [X [Z Y]] becomes[[X Z] Y] prior to inversion of X For more on rebracketing see below


On the other hand string-vacuous (ie noninverting) Local Dislocation is not subject tothese same locality conditions (although of course it affects only string-adjacent elements) Essen-tially string-vacuous lsquolsquorebracketingrsquorsquo is freely permitted much as proposed by Sproat (1985)For example in (12c) X may lsquolsquoescapersquorsquo the constituent C that it was originally peripheral withinsince it will still maintain a left-adjacency relation to Z

To reiterate there is an important difference between Lowering Merger and Local DislocationMerger Lowering is sensitive to syntactic headedness and can therefore affect elements that are notstring adjacent Local Dislocation however is sensitive to relations of adjacency and precedencebetween constituents and not to syntactic headedness directly Thus Local Dislocation mustalways be local as its name suggests it cannot skip any adjoined elements as Lowering canOnly adjacent elements can be reordered by the operation and an intervening (syntactic) adjunctcannot be ignored

The formation of English comparatives and superlatives of the type tall-er tall-est providesa clear case in which a Vocabulary-specific operation is constrained to apply under linear adja-cency To begin with we take the syntactic structure to be one in which the comparative orsuperlative features dominate the position of the adjective (Abney 1987) The realization of thesemorphemes is dependent upon whether or not they combine with the adjective they dominateAs is well known there is a prosodic condition on the host the comparativesuperlative morphemecan combine only with an adjective with one metrical syllable

(14) a John is smart-er than Billb John is mo-re intelligent than Billc John is intelligent-er than Billd John is mo-re smart than Bill

The correlation that we establish here is simple the suffixation of the comparativesuperlativemorpheme is dependent upon the prosodic shape of the host and therefore happens after theinsertion of specific adjectivesThe information that is required for the process to occur is Vocabu-lary specific and because structures are linearized by Vocabulary Insertion the process is definedover a linearized structure Accordingly the comparativesuperlative morpheme cannot appearon the adjective when there is an intervening adverbial this is seen clearly with the superlative1 0

10 The point can also be illustrated with the comparative but with greater difficulty The reason for this is that withthe comparative there are two distinct scopal readings for the adverb one in which it takes scope over the entire compara-tive-adjective and one in which it intervenes between the comparative morpheme and the adjective Thus (i) is grammati-cal but only on a reading in which the adverb takes scope over the comparative

(i) The DuPonts are amazingly t rich-er than the Smiths

That is the degree to which the DuPonts are richer is amazing The syntactic structure is thus one in which the adverbdominates the comparative morpheme This may be compared with a case in which the adverb does not take scope overthe comparative in such cases no combination of adjective and comparative morpheme is possible

(ii) The DuPonts are mo-re amazingly rich than the Gettys

That is both families are amazingly rich but the DuPonts are more so The reason for this is that with the combinedcomparative form the only structure available is the one in which the adverb does not intervene between the position ofthe comparative morpheme and the potential host adjective


(15) a Mary is the mo-st amazingly smart person b Mary is the t amazingly smart-est person

When amazing appears as a modifier of smart it is structurally between the position of thecomparativesuperlative morpheme and the adjective Accordingly it is linearized between thesetwo elements Its presence in this position prohibits superlative -st from being merged with smart(15b) forcing the presence of mo-st

This analysis here treats adjacency as a necessary but not sufficient condition for the forma-tion of a synthetic form That is if the comparativesuperlative morpheme is to combine with theadjective then they must be adjacent But the converse does not hold If the two are linearlyadjacent a synthetic form will not necessarily result Other factors some of them structural willdetermine whether or not Local Dislocation takes place

4 A Model

The device of Morphological Merger as developed by Marantz (1988) captures the generalizationthat clitics are of essentially two types lsquolsquoperipheralrsquorsquo clitics which are either at the edge of amaximal projection or in peninitial or penultimate lsquolsquosecondrsquorsquo position within that phrase andlsquolsquoheadrsquorsquo clitics which adjoin to the head of a phrase1 1 The insight is that the notions lsquolsquoheadrsquorsquoand lsquolsquoperipheryrsquorsquo are derivative of the type of Merger that applies and that the type of Mergerthat applies is different depending on the stage of the PF derivation at which the operation takesplace Figure 1 illustrates the PF branch of the grammar as we envision it The architecture ofthis modelmdashin particular the ordering of events within the derivationmdashmakes particular predic-tions about what can and cannot happen in Morphology In this section we make these predictionsprecise the remainder of the article exemplifies and defends these ideas

41 The Local Dislocation Hypothesis

By the Late Linearization Hypothesis (8) linear order is imposed on a string only at VocabularyInsertion and after linearization Local Dislocation Merger can manipulate only string-adjacentelements From this premise it follows that if a movement operation is sensitive to properties thatare supplied at Vocabulary Insertion it will necessarily apply only to string-adjacent elementsProperties supplied at Vocabulary Insertion include the specific identity of Vocabulary items (iewhether beech has been inserted into a morpheme as opposed to elm or poplar) including anyidiosyncratic properties of the inserted item such as its phonological features or its inflectionalor other diacritical class features If a movement operation is sensitive to such properties of

11 On our view Prosodic Inversion defined by Halpern (1992b) as operating in terms of prosodic subcategorizationwould only apply to sentence-peripheral elements That is if a clitic has a dependency that is purely prosodic it wouldall other things being equal simply lean on a host rather than undergoing Merger Whether or not purely prosodicoperations of this type are necessary beyond the types of Merger that we have proposed is not clear however


(Syntactic derivation)

PFLF BRANCHING

PHONOLOGICALFORM

macr

Lowering

LocalDislocation

(ProsodicInversion)

Building ofprosodic domains

Hierarchical arrangementof morphemes

Linearization imposed byVocabulary Insertion

MORPHOLOGY

VocabularyInsertion

Figure 1The PF branch of the grammar

Vocabulary items we call it Vocabulary sensitive The Local Dislocation Hypothesis can thenbe stated as follows1 2

(16) The Local Dislocation HypothesisIf a movement operation is Vocabulary sensitive it involves only string-adjacent items

For example suppose A B and C are terminal nodes

(17) [A B C] [B CÀ] or [B A`C]

If the movement of A to C depends on the specific Vocabulary items inserted into A or C andnot solely on syntactic or abstract feature properties of A and C then the movement is blockedwhere some other element B intervenes

Effectively the Local Dislocation Hypothesis establishes an important correlation betweenthe locality of a movement operation and what would have been called the lsquolsquoselectionalrsquorsquo propertiesof the clitic in the Lexicalist approach of Klavans (1985 1995) or the Autolexical Syntax theoryof Sadock (1991) Where a clitic demands a host having a particular identity such as inflectionalclass morphological category or phonologicalweight then in our terms the operation is Vocabu-

12 It would also be possible to state a condition like (16) in terms of sensitivity to phonology Thus the generalizationthat is captured here could be captured as well in a theory in which Roots are not inserted late (eg Halle 1990 seeEmbick 2000 and Chomsky 2001 for related discussion) In principle Vocabulary-sensitive and phonology-sensitiveoperations could be distinct but we will not pursue this line of inquiry here


lary sensitive and the clitic and the host must be string adjacent prior to Merger Inversely wherethe clitic element is indifferent to the Vocabulary item that fills its host morpheme but perhapssensitive to the features on a node then it is possible (although not necessary) for the clitic toreach its host by undergoing Raising or Lowering operations that are permitted to cross syntacticadjuncts

42 Late Lowering

While the Local Dislocation Hypothesis is the main focus of the present investigation we willalso point to several other predictions made by the model depicted in figure 1 As shown thereLowering occurs in Morphology after any application of syntactic movement It follows fromthis that while syntactic movement may remove the environment for Lowering to apply theopposite can never hold

(18) The Late Lowering HypothesisAll Lowering in Morphology follows all movement in syntax Lowering can neverremove an environment for syntactic movement

In other words Lowering operates only on the structure that is the output of syntax that is thestructure after all overt syntactic operations have occurred For example English Lowering of Tto v cannot apply where vP has been fronted to a position higher than T (see also Bobaljik 1995)1 3

(19) a Mary said she would quietly play her trumpet and [v P quietly play her trumpet]1

she did t1 b Mary said she would quietly play her trumpet and [v P quietly play-ed2 her trumpet]1

she t2 t1

In (19a) vP fronting makes it impossible for postsyntactic Lowering to move T to v (19b) showswhat would result if Lowering could precede vP fronting We predict that no language shouldpermit lowering of a head Y into an XP that subsequently raises above Y

43 Summary

The stages in the model that we have discussed above are part of a larger conception of the PFbranch which seeks to find the operations that are required in deriving a phonetic expressionfrom a hierarchical arrangement of syntactic terminals The model is assumed to be explicitlytransformational Stages in the derivation involve the addition or manipulation of structures orfeatures that are input from a prior stage Unless otherwise indicated this amounts to the additionof information that is structural distinctions from syntax will still be visible at the stage of thederivation at which for example Vocabulary Insertion and Local Dislocation apply

13 The argument here is valid on the assumption that the do in the clause with ellipsis is not emphatic do


5 Lowering in Bulgarian DPs and the CliticAffix Distinction

Bulgarian shows a lsquolsquosuffixedrsquorsquo definite article here abbreviated Def which has a cliticlikedistribu-tion within the DP (see eg Elson 1976 Scatton 1980 Sadock 1991 Halpern 1992a Tomic1996 Borjars 1998 Caink 2000 Franks 2001) We demonstrate in section 51 that the distributionof Def within the Bulgarian DP is the result of its being the head of D and subject to Loweringas we have defined it above1 4 In section 52 we provide further facts about the Bulgarian Defconcentrating on morphophonology In particular Def shows a number of properties that arecharacteristic of affixes in Lexicalist theories in which affixes and clitics are syntactic and lexicalentities respectively (see eg Zwicky and Pullum 1983 and much subsequent work that assumesthe validity of this distinction) We demonstrate that the distribution of Def has to be stated insyntactic terms and that the syntacticlexical split underlying the cliticaffix distinction is incorrect

51 Lowering of Def

The definiteness element Def in Bulgarian appears suffixed to nominals or when these aremodified by adjectives suffixed to the first adjective in a sequence

(20) a kniga-ta book-DEF

b xubava-ta kniga nice-DEF bookc Adj-Def (Adj) N (general pattern)

The so-called suffixed article contrasts with the overt demonstrative which appears in the expectedplace on the (standard) assumption that the DP has the structure in (22) and that demonstrativesappear in the specifier of DP1 5

(21) tazi knigathis book

In cases with prenominal modifiers such as adjectives the structure of the DP is assumed to bealong the lines proposed by Abney (1987) in which adjectives are heads taking NP argumentsIn this structure shown in (22) the A head is the target of Lowering from D (the operation takesplace on an abstract structure Vocabulary items are shown here for clarity)1 6

14 Arguments against purely movement-based treatments are made by Caink (2000) and Franks (2001) who givevery different treatments of Def elements

15 Typically demonstratives are in complementary distribution with Def Caink (2000) citing Arnaudova (1998)notes that the two can cooccur in colloquial Bulgarian

16 With possessive pronouns Def appears suffixed to the pronominal element

(i) moja-ta xubava knigamy-DEF nice book

We take these possessives to be possessive adjectives and structurally in the same position as the adjective in (22) (seeDelsing 1993)The syntactic distributionof the possessive adjectives puts them structurally higher than other adjectives andthe suffixation of Def to this head follows automatically


(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga

When other elements such as adverbs appear between the NP and D a different patternresults Adverbs may not host Def (23a) nor do they block Merger (23b)

(23) a mnog-t star teatrvery-DEF old theaterlsquothe very old theaterrsquo

b mnogo starij- teatrc dosta glupava-ta zabelezIuml ka (Franks and Holloway King 2000)

quite stupid-DEF remark

The pattern displayed above fits well with the idea of Merger at the level of category (ieLowering) which skips interveningadverbs Def lowers across interveningmaterial to the immedi-ately dominated head The fact that Def does not appear on the adverb is then purely structuralwhen Def undergoes Merger it targets the head of its complement stated in terms of syntacticheadedness The adverb being an adjunct cannot be the target of this operation and structurallyis invisible for it1 7

17 Notice that a syntactic approach to the positioning of Def faces several difficulties The fact that adverbs andadjectives appear together before Def shows that a simple head movement analysis of A- or N-movement to D cannotbe adequate The only alternative would be to raise entire APs On the assumption that AP is the complement of D thismove requires that lower material such as the head noun be scrambled out of the NP complement to AP before the APis raised to the pre-D position This stipulation in syntax is avoided altogether on our Lowering solution

Other attempts to treat the pattern syntactically assuming a different phrase structure for the DP have also beenproposed See Caink 2000 for a critique


52 Bulgarian DPs and the CliticAffix Distinction

We now turn to an illustration of the incorrectness of the distinction between clitics and affixesthat stems from Zwicky and Pullum 1983 and much subsequent work Our treatment stems fromthe discussion but not the analysis in Franks 20001 8 The pattern shown by Def proceeds inconjunction with an examination of possessive clitics As an illustration consider the followingphrase

(24) kniga-ta vibook-DEF yourCL

lsquoyour bookrsquo

Franks argues that while the clitics are syntactic in nature Def is affixal and therefore generatedon its host through a lexical process In our view however this attempt to maintain the modulardistinction between affixation and cliticization misses essential generalizations

Franksrsquos arguments for a lexical treatment of Def involve a variety of phenomena whichin Lexicalist theoretical frameworks have been taken to diagnose lexical as opposed to syntacticderivation One argument for instance is that the phonological form of Def is dependent upona combination of phonological and morphological properties of the host In a Lexicalist modelin which individual lexical items are the initial terminals in a syntactic structure this conditioningof the Def form by its host is problematic unless the two are put together in the Lexicon

In addition Def participates in word-level phonological processes Word-final devoicing isone such process1 9

(25) bratovcIuml ed [bratofcIuml et] lsquocousinrsquomazIuml [masIuml] lsquohusbandrsquo

This process is bled when these words appear with Def

(26) bratovcIuml ed-t [bratofcIuml edt] lsquothe cousinrsquomazIuml -ot [mazIumlot] lsquothe husbandrsquo

On the Lexicalist assumption that this kind of phonological interaction can be found onlyin objects created in the Lexicon Defrsquos behavior forces the conclusion that it is attached via alexical derivation that is that it is an affix

A further argument for the lsquolsquoaffixalrsquorsquo status of Def comes from apparent lexical idiosyncra-sies following the idea that affixation is lexical and therefore subject to unpredictability Theargument is based on the fact that certain kinship terms in Bulgarian have a special appearancein definite form To begin with Bulgarian has a possessive clitic within the DP which is usedonly in definite environments The kinship terms in question show no definite marker

18 Franks bases his analysis on a number of prior discussions of the morphophonologyof Bulgarian definites includingthose of Scatton (1980) Halpern (1992b) and Tomic (1996)

19 A similar argument is adduced from the interaction of liquid metathesisschwa epenthesis and syllabification


(27) a majka mumother hisCL

lsquohis motherrsquob majka-ta mu

(28) a brat otildelsquobrother herCL

lsquoher brotherrsquob brata-at otildelsquo

The syntactic environment is clearly definite however as is clear from cases with adjectives

(29) xubava-ta mu majkapretty-DEF hisCL motherlsquohis pretty motherrsquo

Furthermore other kinship terms do in fact show Def

(30) djado-to mugrandfather-DEF hisCL

lsquohis grandfatherrsquo

The differences here between for example lsquomotherrsquo and lsquograndfatherrsquo are taken to be hallmarksof lexical idiosyncrasy Following the idea that affixes are lexical while clitics are syntactic thispattern is taken as an argument for the affixhood of Def Combined with the prior argumentsthe conclusion is that it must therefore be affixal that is lexically derived

Franks contrasts affixal Def with the possessive clitics seen throughout the prior examplesPhonologically the possessive clitics form a Phonological Word with their hosts to the left butdo not interact with the phonologicalprocess discussed above to the extent that this can be clearlydetermined Thus following the assumptions enforcing the cliticaffix distinction it must be thecase that the clitics are syntactic elements that is not affixes and not lexical However as Franksnotes this leads to an apparent paradox having to do with the distributions of Def and the possess-ive clitics Caink (2000) and Franks (2000) demonstrate clearly that the possessive clitic has adistribution that mirrors precisely the distribution of Def (Def and the possessive clitic italicized)

(31) a kniga-ta vibook-DEF yourlsquoyour bookrsquo

b xubava-ta vi kniganice-DEF your booklsquoyour nice bookrsquo

c mnogo-to ti novi knigimany-DEF your new bookslsquoyour many new booksrsquo


d vecIuml no mlada-ta ni stolicaperpetually young-DEF our capitallsquoour perpetually young capitalrsquo

The problem is that Franksrsquos treatment of the distribution of Def is essentially lexical thatis it appears where it does by virtue of a lexical operation By distributional reasoning the cliticwhich has the same distribution should be treated identically But according to the tenets ofLexicalism this is a contradiction clitics are syntactic and hence cannot be generated lexically

This paradox does not arise in the non-Lexicalist framework outlined above Assume thatCl(itic) is attracted to definite D (see Cardinaletti1998 and Schoorlemmer 1998 for this connectionalbeit in different structural terms) such that syntax outputs a structure in which it is adjoinedto this head (the first tree in (32)) The Lowering operation responsible for the distribution ofDef affects D and will then automatically carry Cl along with Def when it applies When Cl isattached to D[d e f ] Lowering moves the entire D which is internally complex2 0 In the trees in(32) illustrating this process actual Vocabulary items appear in terminal nodes for expositorypurposes only as the movement is Lowering the actual content of these nodes is abstract2 1

(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi

This treatment captures the identical distribution of Def and the possessive clitics directlyD[d e f ] is placed by Lowering with the DP Cl when present is obligatorily attached to D[d e f ] insyntax prior to the stage at which Lowering occurs thus cliticization onto D feeds the laterLowering process

20 See section 6 for structural refinements about the objects affected by Merger21 In cases in which there is an overt demonstrative and a clitic nothing further needs to be said The demonstrative

is in the specifier of DP The clitic has been attracted to D[def] which does not contain Def and hence does not need toundergo Lowering


Franks considers various lsquolsquohybridrsquorsquo treatments of the identity in distribution analyses moti-vated by the idea that Def has to be lexical while clitic placement has to be syntactic There ishowever no need to resort to a hybrid treatment in the first place The facts illustrated aboveprovide further evidence that elements whose distribution or provenance must be essentiallysyntactic can occasionally interact with lsquolsquolexicalrsquorsquo phonological processes (see Embick 1995 forthis type of argument) Additional cases of this type are presented in Hayes 1990 The point issimply that different modes of phonological interaction do not argue for a lexical versus syntacticmode of composition All composition is syntactic and cases like Bulgarian Def make this pointclearly Something must be said about the connection between structures like those resulting fromthe lowering of Def and the word-level phonological interactions illustrated earlier but a treatmentof such facts is beyond the scope of the present discussion

The other properties of Def that were taken as evidence for lsquolsquolexicalrsquorsquo derivation can behandled directly on our treatment The positioning of Def happens prior to Vocabulary InsertionThus what is being lowered is D[d e f ] that is a node with features The fact that the phonologicalform of Def varies according to properties of the host is unsurprising it is merely contextualallomorphy in Defrsquos Spell-Out In addition the apparently exceptional cases of kinship termswith no overt Def could perhaps be treated as involving nouns that take a special zero allomorphof Def This contextual allomorphy for Def occurs only when it is adjoined to one of the nounsin question This accounts straightforwardly for Defrsquos lsquolsquoreappearancersquorsquo when say majka lsquomotherrsquois premodified by an adjective In such cases Def is adjoined to the adjective and spelled outovertly accordingly2 2

53 Conclusions

Within the Bulgarian DP Def lowers to the head of its complement illustrating clearly one ofthe types of PF movement The more general conclusion of the discussion concerns the utilityof the distinction lsquolsquoclitic versus affixrsquorsquo Without the modular distinction lsquolsquolexical versus syntacticrsquorsquothis distinction becomes a matter of terminology without theoretical consequences

6 Refinements of the Model Morphosyntactic Words and Subwords

The preceding sections have illustrated the contrast between Lowering and Local Dislocation andhave shown how the locality conditions of each are identified with distinct stages of the modelof grammar we espouse In this section we refine our definitions of the elements that undergo

22 Two remarks are in order hereFirst the manner in which Def is spelled out based on the host it attaches to has a parallel in English T which is

subject to host-dependent allomorphy when it undergoes Lowering to vSecond Caink (2000) developing the notion of alternative realization found in Emonds 1987 proposes that Def

and the possessive clitic may appear somewhere inside the extended projection in which they occur According to ouraccount in which the clitic is moved to D syntactically and Def undergoes Lowering after syntax the range of placesin which Def could be found is sharply circumscribed it can only appear on the head of Drsquos complement We take thisto be a more restrictive account of postsyntactic movements


Merger We introduce a distinction between two types of objects in Morphology which we callmorphosyntactic words and subwords and demonstrate that movement of each by Merger obeysdistinct locality effects

61 Definitions

First we define the morphosyntactic word as follows2 3

(33) At the input to Morphology a node X0 is (by definition) a morphosyntactic word(MWd) iff X0 is the highest segment of an X0 not contained in another X0

For example consider the structure in (34)

(34) Z adjoined to Y Z+Y adjoined to X

XP

X WP

Y

Z Y

X

hu

gd

ab

In (34) X0 4 Z`Y`X constitutes an MWd but Y0 4 Z`Y does not (at least after adjunctionto X0 ) since Y0 is dominated by X0

Second we define subword as follows

(35) A node X0 is a subword (SWd) if X0 is a terminal node and not an MWd

In (34) Z0 the lower segment of Y0 and the lowest segment of X0 are all SWds Thus an SWdis always a terminal consistingof a feature bundle or if node labels are required an X immediatelydominating a feature bundle and nothing else while being itself dominated In the case in whicha node immediately dominates a single feature bundle this will be by definition an MWd andnot an SWd2 4

In sum any terminal that has undergone head movement in syntax to adjoin to another head

23 This definition seems to correspond to what Chomsky (1995) calls H0max24 In addition intermediate levels of structure such as the Y dominating Z and Y are neither MWds nor SWds

Whether or not such structural units have a particular status in movement operations is an open question we know ofno cases in which they play a role


or any dissociated morpheme adjoined to another head in Morphology will count as an SWdAs we will demonstrate MWds and SWds are the basic atoms of postsyntactic movement opera-tions and these distinct objects impose important restrictions on the types of possible Mergers

62 Two Kinds of Local Dislocation

Both MWds and SWds can be moved by Local Dislocation2 5 We propose that when an MWdis moved by Local Dislocation it adjoins to an adjacent MWd and when an SWd is moved byLocal Dislocation it adjoins to an adjacent SWd This type of restriction mirrors the commonlyheld assumption that syntactic movement treats heads and phrases differently We first reviewan example of each type and then consider the formal properties of Merger that lead to thisrestriction

621 Local Dislocation of Morphological Words The Latin enclitic -que used in conjunctionand roughly equivalent to and appears after the first word of the second of the two conjuncts itappears with

(36) Input (C o n ju n c t 1 X Y) -que (C o n ju n c t 2 W Z)Surface (C o n ju n c t 1 X Y) t (C o n ju n c t 2 W-que Z)

In addition -que which is itself an MWd attaches not to the first SWd of its complement butto the first MWd of its complement (see Marantz 1988)

(37) a [[bonòtilde pueròtilde ] [-que [bonàe puell ae]]] good`NOMPL boy`NOMPL and good`NOMPL girl`NOMPL

b (after Merger) bonòtilde pueròtilde bonàe`que puellàelsquogood boys and good girlsrsquo

c bonòtilde pueròtilde bon-queàe puell ae

Here -que does not interpolate inside the MWd bonàe even though this word is internallycomplex In other words given the structure in (38) X an MWd only (39a) is a legitimatetransformation of (38) (39b) is not

(38) [X [[W W`Y] Z]]

(39) a Possible [[W [W W`Y]`X] Z]b Impossible [[W [W W`X]`Y] Z]

The appearance of -que when it occurs with prepositional phrases adds a slight complicationto this picture With disyllabic prepositions -que typically attaches to the P itself but with manymonosyllabic prepositions it attaches to the complement of P (Ernout and Thomas 1951120)2 6

25 Our proposal does not however allow subconstituents of MWds to move by Raising (Excorporation) or Lowering26 Ernout and Thomas identify exceptions to this pattern which are categorized into a few groups (a) certain fixed

phrases (b) cases in which the complement of the preposition is either a demonstrative or a form of the 3rd personpronoun is and (c) cases in which the preposition is repeated in each conjunct In general the process does not seem tobe obligatory


(40) a i circum-que ea locaaround-and those placeslsquoand around those placesrsquo

ii contra-que legemagainst-and lawlsquoand against the lawrsquo

b i in rebus-quein things-and

lsquoand in thingsrsquoii de provincia-que

from province-andlsquoand from the provincersquo

In such cases a string-vacuous variety of Local Dislocation can apply to phonologically light(monosyllabic) prepositions and their complement uniting them into a single SWd2 7

(41) -que [in re`bus] que [in`re`bus]

Subsequent to this Local Dislocation of -que places it after the P`N unit ignoring the prepositionmuch as the adjectival desinence is ignored in the example in (37)

(42) -que [in`re`bus] [in`re`bus`que]

Examples of this type illustrate two important points First Local Dislocation applies twiceoperating from the more deeply embedded structure outward that is cyclically first the string-vacuous application uniting P and its complement and then the outer application placing -quewith respect to this derived unit Second the MWd status of -que determines where it is placedin a derived object consisting of more than one SWd it is placed after the entire derived MWdnot after the adjacent SWd

622 Local Dislocation of Subwords On the other hand where Local Dislocation targets SWdsit adjoins them to adjacent SWds For example in Huave (Huavean spoken in Mexico) thereflexive affix -ay appears directly before the final inflectional affix of a verb if any Considerthe following examples (Stairs and Hollenbach 1981 reflexive affix italicized)

(43) a s-a-kohcIuml -ay1-TH-cut-REFL

lsquoI cut myselfrsquob s-a-kohcIuml -ay-on

1-TH-cut-REFL-PL

lsquowe cut ourselvesrsquo

27 Support for this position which makes these prepositions proclitic on their complements is found in Latin orthogra-phy which occasionally treated monosyllabic prepositions as part of the same word as the complement (see Sommer1914294 Leumann Hofmann and Szantyr 1963241)


(44) a t-e-kohcIuml -ay-osPAST-TH-cut-REFL-1lsquoI cut (past) myselfrsquo

b t-e-kohcIuml -as-ayPAST-TH-cut-1-REFL

(45) a t-e-kohcIuml -as-ay-onPAST-TH-cut-1-REFL-PL

lsquowe cut (past) ourselvesrsquob t-e-kohcIuml -ay-os-on

PAST-TH-cut-REFL-1-PL

Although -ay lsquoreflexiversquo directly follows the Root and precedes -Vs lsquo1st personrsquo in (44a) itfollows -Vs lsquo1st personrsquo in (45a)2 8 We can account for these facts by assuming that -ay isstructurally peripheral to the verbìnflection complex but undergoes Local Dislocation to left-adjoin to the rightmost inflectional affix

(46) a [[s-a-kohcIuml ]on]ay [[s-a-kohcIuml ]ayòn]b [[[s-a-kohcIuml ]as]on]ay [[[s-a-kohcIuml ]as]ayòn]

623 Discussion The Huave examples (43)ndash(45) establish that Local Dislocation cannot berestricted to MWds Dislocation of SWds must also be permitted But movement of an MWd toadjoin to an SWd or of an SWd to adjoin to an MWd as in (39b) is apparently impossible Toobtain the correct restriction on inversion operations it suffices to require that

(47) If a Merger operation moves an element A to a target B then A and the head of B areeither both MWds or both SWds

Clearly the same restriction is standardly assumed to hold when the operation in question issyntactic movement XPs adjoin to XPs and X0 s adjoin to X0 s (see Baltin 1991 for some discus-sion)

For the purposes of exposition it will be convenient to distinguish notationally betweeninstances of Merger of an MWd and Merger of an SWd To make explicit whether a particularinstance of Local Dislocation moves an MWd or an SWd we will use the symbol Aring to denoteadjunctionof one SWd to another The symbol ` will continue to denote other types of adjunctionas is standard

The theory predicts certain interactions concerning domains that are accessible for Mergeroperations and concerning the relative ordering of operations First a complex X0 created insyntax (or by Lowering) cannot be infixed within another X0 during Morphology In other wordslsquolsquosecondrsquorsquo position for an MWd is after (or before) the first (last) MWd in a phrase and nowhereelse2 9

28 The 1st person suffix shows an alternation -as-os-is29 Klavans (1995) and Halpern (1992b) (among others) discuss cases in which second position appears to be either

after the first word or after the first phrase On the present proposal positioning after the first phrase cannot arise from


(48) X [Y 0 Y Z] [Y 0 Y`X Z] where X is an MWd

Second an SWd (ie a terminal node within a complex X0 created by Raising or throughthe insertion of dissociated morphemes in Morphology) can never adjoin to an element outsidethat X0 Schematically

(49) X [Y 0 Y Z] Y`X [Y 0 Z] where X is an MWd and Y is an SWd

The latter point is of particular interest in light of the analysis of Bulgarian Def aboveLowering in the Bulgarian DP moves an MWd D to the head of its complement When a clitichas attached to the D in the syntax Def cannot be lowered independently of the clitic as thiswould be a case of lowering an SWd to an MWd which is prohibited on our approach Theprocess thus affects the MWd D dominating both [Def] features and Cl with the result that thepossessive clitic and Def have an identical distribution

63 The Lithuanian Reflexive

The behavior of the reflexive morpheme -si in Standard Lithuanian (Senn 1966 Nevis and Joseph1992) provides an important showcase for the interaction of Local Dislocations at the SWd levelIn simple verbs such as (50a) -si appears as a suffix to the complex of verb stem ` tense andagreement inflection (50b) In verbs with certain lsquolsquopreverbrsquorsquo prefixes historically derived fromadverbs -si appears not after the verb ` inflection but between the prefix and the verb (51b)

(50) a laikau lsquoI consider maintainrsquob laikau-si lsquoI get alongrsquo

(51) a isIuml -laikau lsquoI preserve withstandrsquob isIuml -si-laikau lsquoI hold my standrsquo

When two such prefixes appear before the verb the reflexive morpheme appears betweenthem

(52) a pa-zIuml inti lsquoto know [someone] to recognizersquob su-si-pa-zIuml inti lsquoto become acquainted withrsquo

Furthermore when a verb is negated by the prefix ne- the negation prefix appears before anypreverbs and reflexive -si is immediately to its right (examples from Dambriunas Klimas andSchmalstieg 1972)

(53) a alsquosIuml lenkiulsquo lsquoI bendrsquob alsquosIuml lenkiuo-si lsquoI bowrsquo (lit lsquoI bend myselfrsquo)c alsquosIuml ne-si-lenkiulsquo lsquoI do not bowrsquo

Local Dislocation It must be the case that the initial XP in question has raised (eg by a topicalization fronting) tosentence-initial position Whether or not a prosodic operation is needed in addition to movement for cases in which anXP precedes a clitic is an open question


The generalization emerging from these data is that -si appears as a suffix to the first prefix(of a certain type) on the verb where there is no prefix -si suffixes to the verb ` inflection Interms of Local Dislocation -si has moved from a position as a prefix to the verb complex tosecond position where crucially the verb ` inflection together count as a single lsquolsquopositionrsquorsquoLike Latin -que Lithuanian -si cannot be positioned between the stem and an inflectional suffixbut unlike -que -si can be positioned inside an MWd namely between a prefix and a followingstem or prefix

Our analysis of these facts is as follows We assume that in Lithuanian Pr(everb)s are adjoinedto V and this complex moves to Neg if present and further to T

(54) [T P [Neg1`[Pr`Pr`V]2 ]`T [N e g P t1 [V P t2 ]]]

We take -si to be a dissociated morpheme inserted in Morphology left-adjoined to the highestsegment of the MWd of which v is a member that is to the entire [ V]`T complex in thecases under discussion3 0 From this position it undergoes Local Dislocation As an SWd and notan MWd itself -si trades its relation of left-adjacency to the [ V]`T complex for a relationof (right-)adjunction to the left-peripheral SWd within this complex namely the leftmost prefix3 1

(55) [-si [Pr V T]] [[Pr Aring si V T]]

This procedure gives the correct results whenever the verb has a prefix or is negated How-ever since both V and (the lower segment of) T are SWds we incorrectly predict that -si willdislocate to between V and T in a verb with neither negation nor a prefix

(56) [-si [V T]] [[V Aring si T]]

The reason this does not happen evidently is that V and T form a unit impenetrable to LocalDislocation This fact reflects another namely that in most Indo-European languages suffixesform closer phonological domains with stems than do prefixes prefixes are more lsquolsquolooselyrsquorsquoattached than suffixes and suffixes show more pronounced allomorphy conditioned by the stem

30 The si under discussion here is classified as lsquolsquoreflexiversquorsquo for convenience only In fact it appears in a number ofdifferent verbal types in Lithuanian many of which are not actually reflexive This pattern is typical of voice morphologythat does not actually instantiate a syntactic terminal although we cannot undertake a detailed analysis showing that thisis the case here (see Embick 1997) For full distribution of this element in Lithuanian see GeniusIuml ieneCcedil 1987

The statement of the process in terms of lsquolsquohighest MWdrsquorsquo is intended to cover cases in which the v has combinedwith Asp but not with T that is participles The pattern of -si in these cases parallels its pattern in the domain of tensedverbs

31 The position of -si is not constant across Lithuanian dialects Endzelotilde ns (1971) provides data from a dialect inwhich -si is suffixed to the verbìnflection complex even when there is a preverb

(i) su-pranta-siPR-understand-REFL

lsquo(they) understand each otherrsquo

In such dialects -si is presumably right-adjoined to (the highest segment of) T0 and does not undergo any dislocationDoubling of reflexives also takes place in certain dialects Senn (1966sec 401) reports that in Lower Lithuanian (Nieder-litauisch) s(i) appears after the verb when negation is present but after verbal prefixes when these are present in thelatter cases doubling is also possible with a second s(i) appearing after the verb We do not attempt a full analysis ofsuch forms here but see section 71 for some discussion of doubling


than do prefixes3 2 Where the suffix in question is T this closer phonological affinity to the stemis directly at odds with the syntactic derivation where by hypothesis T has attached last duringsyntactic head raising (54)

To express this restructuring we propose that T in Lithuanian always undergoes string-vacuous Local Dislocation adjoining to its left neighbor V

(57) [V T] [[V 0 V Aring T]]

Recall now that Local Dislocation manipulates SWds where these are defined as the terminalelements within an MWd Because SWd status is defined before Local Dislocation if the SWdT0 Aring -adjoins to the SWd V0 as in (57) the result is a single complex SWd and not two SWdsIt follows that subsequent Local Dislocation will treat the [V Aring T] unit as a single lsquolsquopositionrsquorsquoThis is indeed what we find since -si dislocates to the left of [V Aring T] as a whole

The ordering of the two hypothesized Local Dislocations is not arbitrary it is predicted bythe principle of cyclic application Since -si is adjoined to T and T is adjoined to [ V] thenany dislocation operating over T and V will precede any dislocation operating over -si and T3 3

7 Interactions and Requirements

71 Swedish Determiners and Definiteness Marking

The distributionof definiteness marking in the Swedish DP providesa case study for the interactionof requirements movement operations and support processes There is an extensive literature onthe topic for relevant references see Borjars 1998

A definiteness marker in Swedish appears suffixed to the noun when there is nothing elsein the DP

(58) mus-enmouse-DEF

lsquothe mousersquo

Overt determiners cooccur with definiteness marking on nominals when for instance anadjective precedes the head noun resulting in a type of doubling3 4

32 Evidently the formation of definite adjectives in Old Lithuanian shows a similar pattern with respect to the statusof prefixes An element that was historically pronominal (the iAEligo-stem pronoun) appears between prefixes attached todefinite adjectives and the stem although some doubling of this element in postadjectival position seems to occur aswell See Stang 196670 and ZinkevicIuml ius 19577ff for discussion

33 See Minkoff 1994 for a discussion of cyclicity and Merger in the case of Caribbean Spanish clitics34 A different situation obtains with overt determinersdemonstratives Certain determiners require the presence of

the definite marker while others are unable to cooccur with it (no syntacticosemantic difference is known to distinguishthe determiners that do and do not require lsquolsquodouble definitenessrsquorsquo we thank Anders Holmberg for discussion of thispoint)

(i) Swedish Det and Defa den gamla mus-en mus lsquothethat old mouse-DEFrsquob den mus-en mus lsquothat mousersquoc den har mus-en mus lsquothis mousersquod denna mus mus-en lsquothis mousersquo


(59) den gamla mus-enthe old mouse-DEF

lsquothe old mousersquo

The apparent lsquolsquodoublingrsquorsquo in which an overt determiner den cooccurs with a definiteness-marked noun is obligatory Structurally the head noun is crucially at issue here That is it isnot simply the case that a [def] element is phonologically associated with the right of an entireDP as for instance the English possessive is (eg [the guy on the left]rsquos dog) This is clear inexamples with postnominal modification (Borjars 1998)

(60) Gris-en med la ng svans grymtadepig-DEF with long tail gruntedlsquoThe pig with the long tail gruntedrsquo

Thus there is always definiteness marking on the head noun whether or not prenominalmodifiers appear The fact that definiteness marking cooccurs with a number of overt determinerssuggests that the definiteness marker on nouns in such cases is a dissociated morpheme ratherthan the realization of D serving as a sort of agreement Notice that the alternative to this is tocomplicate syntax unnecessarily holding for instance that Swedish has two distinct D heads undercertain circumstances but not others (Kester (1992) makes such a proposal see the critique inDelsing 1993)

In terms of where the requirements responsible for the attested pattern are localized Swedishprovides an instance in which two elements D[d e f ] and the head N force certain conditions to bemet As noted above a head N in a definite DP in Swedish always shows marking for definitenessOur interpretation of this is as follows First Swedish has a condition to the effect that N in thecontext of D[d e f ] must be marked morphologically for definiteness in a kind of concord Thisamounts to a well-formedness condition on N At the same time there is also a morphophonologicalrequirement on the syntactically projected D[d e f ] it must have a host Thus when D[d e f ] and thehead N cannot be put together den instantiates the D position (see Santelmann 1992) The tworequirements are stated as follows

(61) Requirements (imposed at PF)a The head N must be marked with definiteness when D is [def]b D[d e f ] must have a host

It is possible for the overt determiner to appear with a definiteness-marked noun without an adjective as in denmus-en the interpretation is that of a demonstrative

A further quirk involves restrictive relative clauses Borjars (1998) reports optionality here either a suffixed nounor an overt determiner with a suffixed noun (with nonrestrictive relatives only the suffixed-noun version is possible)

(i) a mus-en som inte mouse-DEF that not

b den mus som inte DET mouse that not


Notice that in requirement (61a) the condition may be triggered either by D[ d e f ] itself orby virtue of the presence of one of the other demonstrative elements noted above that requiredefiniteness

These requirements are directly affected by what has taken place in the syntactic derivationIn the case in which nouns without preceding modifiers appear suffixed with the definitenessmarker the simplest analysis is to invoke head movement to D in this we follow Taraldsen(1990) and Delsing (1993) That is3 5

(62) N moves to D if possible

When N-to-D movement occurs each of the requirements in (61) is satisfied The N ismarked with definiteness and D[d e f ] has a host the incorporated N At the same time however(62) is restricted to N and will not occur if N is dominated by a modifier3 6 When syntacticmovement does not occur the two requirements in (61) have not been met and further PFprocesses must apply In such cases we take the output of syntax for an example like (59) to be(63) where A and N Roots are inserted for clarity

(63) Output of syntax

DP

D

D AP

[def] A NP

NGAMLIuml sect sect sect

MUSIuml sect sect sect

35 In line with a specific treatment of derivational morphology within Distributed Morphology we take N here asshorthand for a functional head n analogue of the v of the verbal system (see Marantz 2000)

36 Sandstrom and Holmberg (1994) note an interesting pattern in northern Swedish in which adjectives and nounstogether in a type of compound structure appear suffixed as a whole with the definiteness marker for example comparenorthern Swedish gamm-svart-kjol-n lsquoold-black-skirt-DEFrsquo with Standard Swedish den gamla svarta kjol-en


(64) Processes

DP

D

D AP

[def]

[def]

d- A NP



As we discussed at the outset the assignment of dissociated morphemes occurs at PF inaccordance with language-specific requirements In the present case requirement (61) triggers theassignment of a dissociated morpheme to N in what could be viewed as a type of concord That ismuch as features of say subject DPs are copied onto Agr nodes on T for subject-verb agreement ormuch as features in a DP like [number] appear on A nodes within a DP Swedish has a requirementto the effect that Ns in [def] environments reflect the [def] feature via concord The process thatbrings about this lsquolsquoagreementrsquorsquo is as follows

(65) Assign [def] to the head N in a DP with the [def] property

This process takes place in PF that is after syntax Moreover it applies only when necessaryIn cases in which N moves to D syntax will output a structure in which [N Def] are a unit Thiswill meet requirement (61a) Thus (65) will not apply Syntax therefore can bleed the applicationof the PF process Syntax may also satisfy the second requirement (61b) The other componentof the Swedish determiner system involves the support of stranded D[d e f ] requirement (61b) Asnoted above we follow Santelmann (1992) in regarding this as a case of support by d- of strandedD[d e f ] Once again this process may be bled if N-to-D movement has occurred in syntax

The first point illustrated by this case study concerns the multiple requirements that may beinvolved in PF well-formedness and their interaction with the syntactic derivation Because ofthe existence of requirement (61a) in Swedish morphology we find the doubling of a head thatis relevant to LF interpretation but there is no doubling at the syntacticosemantic level becausethe feature [def] is only copied in PF The interaction of requirements here also paves the wayfor our study of the complicated array of syntactic derivationsmorphologicaland morphosyntacticrequirements rescue operations and Lowering movement that must be acknowledged in theEnglish verbal system the topic of section 72


The second point is that our analysis captures the essential properties of the Swedish DPwithout introducing a complication into the syntactic component The importance of this positionis clear when we look at additional patterns from the Scandinavian languages The situation inSwedish forms a minimal contrast with the distribution of definiteness marking in Danish Thearticle in Danish (definite or indefinite) occurs in complementary distribution with all overt deter-miners when it appears suffixed to a noun

(66) Danisha mand-en lsquoman-DEFrsquob den unge mandc den unge mand-en lsquothe young manrsquo

We take the difference between Danish and Swedish to be localized in Morphology Specifi-cally Danish has requirement (61b) but not (61a) It therefore requires D[ d e f ] to be supportedbut it crucially does not require the type of agreement that results in the doubling of [def] inSwedish In terms of our analysis Swedish and Danish are quite similar syntactically the differ-ence between them is found at PF The alternative to this which is to say that Swedish and Danishdiffer significantly in terms of where D is located in the DP and in terms of the types of movementoperations they allow complicates the syntax unnecessarily in comparison with our treatmentwhich analyzes the phenomena as arising from the independent requirements of syntax and PF

72 T and v in English

Earlier we noted following previous treatments that T attaches to V in English by LoweringWe now turn to a further questionconcerning the nature of the requirements surroundingLoweringin English concentrating in particular on the interaction of Lowering with do-support

The first point we consider is one discussed by Bobaljik (1995) the Lowering of T to Vtreats intervening adverbs as invisible3 7 In order to establish this point clearly however we mustaddress a specific point concerning the relative positions of T V and intervening adverbs Thedistribution of certain adverbs with respect to TP and VP varies as shown by the followingexamples

(67) a John always has read a lot of booksb John has always read a lot of books

In cases in which always is supposed to be invisible for Lowering it cannot be known

37 Bobaljik argues that adjacency is relevant for the process that we treat as Lowering here The major differencebetween the two approaches is that by defining Lowering structurally we have a clear reason why adverbs do not interferewith the process Lowering relates T with the head that it immediately dominates v Other material that intervenes linearlybetween these two positions simply does not count for the process unless there is additional syntactic structure (ieNegP) Note that this difference holds only if we assume that adverbs do not always appear in the specifier position ofaccompanying functional heads If this were the case it would be necessary to stipulate a reason why the heads in questiondo not count for the purposes of Lowering


without further argument whether it is adjoined to TP instead of VP That is the structure of(68a) with Lowering might be that in (68b)

(68) a John always readsb John [T P always [T P T [V P reads]]]

We believe that the invisibility of adverbs for Lowering can be established by adverb phrasesthat can appear only on VP not on TP and that do not block Lowering3 8 Taking the presenceof an overt auxiliary to show the position of T the following examples illustrate this pattern withthe adverb completely

(69) a John has completely destroyed the oppositionb John completely has destroyed the oppositionc John t completely destroy-ed the opposition

Lowering in English is driven by a morphological requirement on finite T Structurally theinvisibility of adverbs is expected given a Lowering analysis of T-to-V movement To this pointthen the English system is unremarkable However examples involving constituent negationshow that the relationship between Lowering and do-support in English is more complicatedthan the analysis above would indicate3 9 Consider first the following examples with constituentnegation sentential negation is provided in (70b) for clarity

(70) a John can always not agreeb John canrsquot always not agree

The relevant examples show that in cases without the modal there is no grammatical declara-tive version either with Lowering or with do-support

(71) a John always not agreesb John does always not agree

The generalizations arising from these examples are that Lowering is impossible and thatdo-support fails as well if T remains in situ For instance when T-to-C movement occurs it isin fact possible to realize T with do

(72) Did he always not agree

What this shows is that when T is removed from the structural domain in which Loweringoccurs do-support is not blocked Similarly when Lowering is precluded by the presence ofnormal negation do-support is once again possible

(73) He didnrsquot always not agree

These examples are crucial for understanding the factors that interact in English to yield do-support phenomena Before this can be seen however the basic elements of do-support must bepresented At first glance this appears to be a morphophonological rescue strategy that is an

38 Additional diachronic evidence for the invisibility is presented in Kroch 198939 We thank Tony Kroch for bringing these cases to our attention


operation aimed solely at resolving the dependencyof an otherwise stranded T Two considerationssuggest that this is not correct The first is based on the examples with constituent negation aboveThey show that do-support cannot apply when constituent negation is present If do-support weresimply the insertion of a dummy stem in Morphology this fact would be entirely unexplainedthere is no reason why the presence of constituent negation should preclude the insertion of doThe second argument is based upon the distribution of features and feature types in the grammarWe assume that do as a light verb is essentially the default Spell-Out of the head v Viewedfrom this perspective in order for do-support to take place there must be a position into whichthe forms of do are inserted that is a v however the cases in which do-support applies in Englishare those in which v cannot combine with T Thus do-support must involve two steps first theaddition of v to T and second the realization of this v as do This is the basis for the secondargument against treating do-support as purely morphological The head v is a syntacticosemanticobject On the assumption that Morphology interprets the output of syntax v is simply not thetype of object that Morphology can insert A consequence of Late Insertion called Feature Dis-jointness in Embick 1997 2000 is that purely morphologicalphonological features are not presentin syntax and syntacticosemantic features are not introduced in Morphology Reintroducing a vin Morphology as is required by the morphological support analysis of dummy do contradictsthis basic and architecturally motivated position concerning features in the grammar4 0

The conclusion drawn from these arguments is that do-support must be essentially syntacticWe therefore propose to treat do-support as occurring in syntax that is there is a specific structuralenvironment in which T is provided with v in syntax4 1 The motivation for this is a syntacticlocality condition governing the relationship between T and v

(74) T must be in an immediately local relationship with v

Here we take immediately local to mean that T must either have a vP sister or be in an MWdwith v This amounts to a strict selectional requirement on T to the effect that it must be localto v So for instance when auxiliaries move to T the requirement is satisfied because T is in acomplex head with v This is a manifestation of a more general principle according to which arequirement of this type can be met by a head X or the XP projected from X in a theory inwhich notions like lsquolsquoXrsquorsquo and lsquolsquoXPrsquorsquo are derivative of structure

If at any stage in the derivation the requirement in (74) cannot be met a syntactic processmust occur to meet the locality condition on T4 2

(75) v is syntactically merged onto T when T does not have a vP complement

This covers two cases simple negation in which T immediately dominates NegP or more pre-cisely S P in the sense of Laka (1990) and T-to-C movement environments in which T has

40 For potential complications see the discussion in Embick 200041 The case of support with d- in Swedish could also be examined in greater detail in light of our claim that some

support phenomena are syntactic42 In cases in which auxiliaries move to T via Neg as in John isnrsquot a chef Trsquos requirement would have to be met

by head movement rather than by do-support


moved In the former case the head S actually hosts either affirmativeemphatic features or [neg]in which case it is referred to as NegP In either case however do-support is found Negationis illustrated above with the emphatic S do-support is also obligatory We take so in exampleslike the following to be realized in the S projection

(76) a John did eat the apples (emphatic)b John did SO eat the applesc John SO ate the apples

When there is no S P or no T-to-C movement there is no do-support The combination ofT and v then comes about in Morphology via Lowering The derivation is sketched in trees (77)and (78) When T is not in an immediately local relationship with v the requirement in (74) mustbe syntactically satisfied by merging a v with T4 3

(77) T does not dominate v requirement of T not met

T

T NegP

Neg

[neg] vP

eat the apples

(78) Do-support in the syntax default v satisfies requirement

T

T

Tv

NegP

Neg

[neg] vP

agree with Bill

43 Stated derivationally in cases in which there is T-to-C movement or S P but there is no auxiliary verb or vattached to T the derivation crashes


The remaining complication to this picture concerns the ungrammaticality of both do-supportand Lowering in sentences with constituent negation We propose to account for this case ofineffability in structural terms In outline the idea is that the syntactic requirement of T is metwhen constituent negation modifies the vP that is T continues to dominate the vP unlike whathappens when a full NegP is present Following the (licit) syntactic derivation Morphology isthen provided with a case in which Lowering must occur to place T on v However whenconstituent negation is present the structural head of Trsquos complement is no longer v rather itis Neg headed by constituent negation The postsyntactic derivation cannot yield an acceptableoutcome via Lowering in this case And because do-support is syntactic as demonstrated abovethere is no morphophonological rescue strategy

Constituent negation is a head adjoined to a phrase in the examples we are considering Ingeneral we take constituent negation to be essentially the same [neg] feature that is found withnormal negation In the lsquolsquonormalrsquorsquo case [neg] is a feature on S and the S P is referred to as NegPBeyond this however Neg can adjoin to other syntactic elements When Neg is adjoined tosome phrase (or head see below) and does not head S it is called constituent negation Butdespite the difference in syntactic distribution the same feature underlies so-called normalnegation4 4

A further question regards what precisely it means for constituent negation to be a headadjoined to a phrase We will explicate this within a version of bare phrase structure (Chomsky1994) When Neg is adjoined to a phrase it does not project further Thus it is potentiallysimultaneously maximal and minimal a possibility discussed in some detail in Chomsky 1995In order to show how this structure for constituent negation interacts with Lowering we mustfirst examine the nature of the Lowering operation in greater detail On the definition we havebeen employing Lowering relates a head to the head of its complement In the cases we haveexamined in previous sections the XP into which Lowering occurs is headed by an MWd X asin (79)

(79) [X P (ZP) [X X (YP)]]

When constituent negation occurs Neg is adjoined to the vP As noted as a head adjoined to aphrase it has the properties of both a head and a phrase The fact that it counts as a minimalprojection (ie an X0 ) in the relevant configuration is crucial The vP that is the complement ofthe T to be lowered has the following structure

(80) [v P DP [v P Neg [ v v Iuml P]]]

In the typical case illustrated in (79) the closest MWd of the XP is identical to the syntactic

44 This accounts for the semantic similarities of the two and for the fact that both lsquolsquonormalrsquorsquo and lsquolsquoconstituentrsquorsquonegation can license negative polarity items given the correct structure


head of the projection X By closest MWd here we mean the structurally highest MWd attachedwithin the complement of the Lowering head In (79) and (80) no head Y within ZP or DPwill be structurally close in this sense as it is embedded within an adjoined projection that is(unambiguously) phrasal4 5 The intuition we will pursue is that Lowering in fact targets not thehead of its complement but the closest MWd (as defined above) of the complement Under normalcircumstances lsquolsquohead of complementrsquorsquo and lsquolsquoclosest MWd of complementrsquorsquo will pick out the sameobject as in (79) In (80) however there is an MWd Neg that is hierarchically higher than thehead v such that the MWd targeted by Lowering (Neg) and the MWd heading the projection (v)are different heads In particular with constituent Neg adjoined to vP the closest MWd in thecomplement of T is Neg itself and this prohibits successful Lowering We illustrate this analysisin greater detail in comparison with adverbs like completely

In the cases under consideration Neg is adjoined to the vP However it is not adjoined asan Adverb Phrase (AdvP) would be AdvPs do not block Lowering or lead to a crash This isstructural Adverbs such as completely appear in AdvPs that are adjoined to the vP (the subjectDP is omitted here for simplicity)

(81) Invisibility of the adverb

T

vPT

AdvP

v

v

Iuml P

v

completely

t with BillAGREEIuml sect sect sect

In this case the structural head of Trsquos complement is v and Lowering applies to yield thegrammatical outcome In the case of constituent negation it is simply the head containing [neg]that is adjoined to the phrase Tree (82) shows [neg]rsquos adjunction to the vP

45 This type of restriction parallels the case of head movement which allows movement to an immediately dominatinghead not for example movement from an adjoined XP to a higher head outside that XP


(82) Constituent negation syntactic requirement of T met

T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not

Unlike in (78) in which T has a NegP complement and thus has a default v merged withit in (82) T still has a vP complement Thus do-support (ie merging default v with T) is nottriggered The structure is thus legitimate and is shipped to the PF component as is At PF T isnot joined with v with the result that Lowering must apply In this case however the targetvisible for Lowering is Neg not v as Neg is the closest MWd of Trsquos complement Thus Loweringcannot apply or if it applies blindly it yields the morphologically illegitimate object Neg-T Oneither scenario the result is ungrammatical In this way the analysis captures the intuition that theineffability of these examples with constituent negation follows from the distinct requirementsof two modules of the grammar4 6

Part of the solution to the constituent negation problem involves the idea that this elementis a head adjoined to a phrase Our account thus predicts that the attachment of a Neg elementinto the MWd in which the verb appears should not create a problem As the facts show thecases with constituent negation and the structural anomaly that it induces contrast sharply withexamples involving prefixes like dis-

(83) John t quickly dis-agree-d with Bill

This follows directly from structural differences between constituent negation and dis- We takethe latter to attach directly to the Root which in the verbal example then moves to v as a unitas shown in (84)4 7

46 An alternative to the treatment proposed in the text is found in Flagg 2000 The analysis ties the crash in questiondirectly to the effects of phases in a cyclic derivation but is similar in other respects to our treatment While the structure-based and phase-based treatments make different predictions we will not explore these differences here

Ultimately the predictive power of the approach developed in the text will have to be examined with further casestudies of Lowering

47 We indicate the features underlying dis- here as X as they are not identical to Neg


(84) dis-agree

v

v

X

AGREEIuml sect sect sectdis

Iuml P

Iuml P

In this structure the head of Trsquos complement is the (internally complex) v which is the target forLowering The grammaticality of (83) follows because the target of T is a legitimate morphologi-cal host for it4 8

The case studies presented in this section highlight the interaction of syntactic and postsyntac-tic movement processes with support phenomena and illustrate the combination of syntactic andmorphological requirements underlying the Swedish determiner system and the English tensesystem In particular we provided an account of do-support according to which this process issyntactic allowing the ineffability found in cases with constituent negation to be accounted fordirectly

8 Conclusions

This article is intended as a contribution to a general theory of the Morphology componentwhich provides phonological expression to abstract syntactic structures We have assumed thatMorphology has recourse to a restricted range of readjustment operations modifying syntacticstructures where such readjustments in some cases lead to surface effects commonly termedlsquolsquomorphosyntactic mismatchesrsquorsquo Included in these readjustments is a mechanism for movementidentified here with MorphologicalMerger as introduced by Marantz (1988) Our proposals basedon case studies from a range of languages make explicit these operations and the primitives ofconstituent structure that they manipulate substantially extending earlier formulations of Merger

First we have identified two species of Merger Lowering and Local Dislocation and haveshown that where the conditions on the operation of the Merger make reference to particularVocabulary items the operation is purely local operating under string adjacency only Thisprediction the Local Dislocation Hypothesis follows automatically from the architecture of gram-

48 Something similar appears to be marginally possible for some speakers with constituent negation replacing dis-The affixation of Neg inside the MWd of the V can be diagnosed morphophonologically

(i) a not agreeb not agrelsquoe

In (ia) not is attached to the vP while in (ib) it is attached to the MWd containing the verb much as is dis- the phonologyis more characteristic of a compound in (ib) than in (ia) In (ib) Neg attaches very low perhaps in the same way thatthe feature realized by dis- is adjoined to the Root


mar proposed here and we believe represents a novel empirical observation that alternativemodels of grammar cannot encode without stipulation Second we have formally defined thecategories morphologicalword and subword as the primitive constituentsof Morphologicalmove-ment operations Further constraints on potential syntaxmorphology mismatches are defined interms of these fundamental categories The distinction between clitics and affixes on the otherhand has no place in our non-Lexicalist approach we have shown that this cumbersome artifactof Lexicalism obscures rather than clarifies basic aspects of the syntax-morphology interfaceThird we illustrated the interaction of postsyntactic requirements with syntactic movement andconcord phenomena In clarifying how multiple requirements across two distinct components ofthe grammar can result in complex surface patterns we have also introduced a novel treatmentof English do-support

More generally the interactions that we have analyzed highlight how our approach differsfrom other attempts to characterize syntax and its relationship to morphology In each of the casestudies presented above the analysis of morphological operations relies crucially on a syntacticstructure The intimate connection between syntax and morphology that underlies our approachis a basic assumption of the Distributed Morphology research program The well-formedness ofany particular surface string results from diverse requirements that are distributed across differentparts of the grammar In this way we reject the view that morphology is insulated from syntaxor can be understood without reference to the grammar as a whole Our proposals instead providea specific framework in which deviations from a direct reflection of syntax in morphology canbe isolated and understood

References

Abney Steven 1987 The English noun phrase in its sententialaspectDoctoraldissertationMIT CambridgeMass

Arnaudova Olga 1998 Demonstratives and the structure of the Bulgarian DP In Papers from the SecondConference on Formal Approaches to South Slavic Languages 1ndash32 Trondheim Working Papersin Linguistics 31 Department of Linguistics University of Trondheim

Baltin Mark 1991 Head movement and Logical Form Linguistic Inquiry 22225ndash250Bobaljik Jonathan David 1994 What does adjacency do In MIT working papers in linguistics 21 The

morphology-syntax connection 1ndash32 MITWPL Department of Linguistics and Philosophy MITCambridge Mass

Bobaljik Jonathan David 1995 Morphosyntax On the syntax of verbal inflection Doctoral dissertationMIT Cambridge Mass

Bonet Eulalsquolia 1991 Morphology after syntax Pronominal clitics in Romance Doctoral dissertation MITCambridge Mass

Borjars Kersti 1998 Feature distribution in Swedish noun phrases Oxford BlackwellCaink Andrew 2000 In favour of a lsquolsquoclitic clusterrsquorsquo in the Bulgarian and Macedonian DP In Papers from

the Third Conference on Formal Approaches to South Slavic and Balkan Languages 170ndash182Trondheim Working Papers in Linguistics 34 Department of Linguistics University of Trondheim

Cardinaletti Anna 1998 On the deficientstrong opposition in possessive systems In Possessorspredicatesand movement in the Determiner Phrase ed Artemis Alexiadou and Chris Wilder 17ndash54 Amster-dam John Benjamins


Chomsky Noam 1994 Bare phrase structure MIT Occasional Papers in Linguistics 5 MITWPL Depart-ment of Linguistics and Philosophy MIT Cambridge Mass [Published in Government and BindingTheory and the Minimalist Program ed Gert Webelhuth 383ndash439 Oxford Blackwell (1995)]

Chomsky Noam 1995 The Minimalist Program Cambridge Mass MIT PressChomsky Noam 2000 Minimalist inquiries The framework In Step by step Essays on minimalist syntax

in honor of Howard Lasnik ed Roger Martin David Michaels and Juan Uriagereka 89ndash155CambridgeMass MIT Press [Also availableas MIT OccasionalPapers in Linguistics15 MITWPLDepartment of Linguistics and Philosophy MIT Cambridge Mass]

Chomsky Noam 2001 Derivation by phase In Ken Hale A life in language ed Michael Kenstowicz1ndash52 Cambridge Mass MIT Press

Dambriunas Leonardas Antanas Klimas and William R Schmalstieg 1972 Introductionto modern Lithua-nian New York Franciscan Fathers

Delsing Lars-Olof 1993 The internal structure of noun phrases in Scandinavian languages Doctoral disser-tation University of Lund

Elson Mark 1976 The definite article in Bulgarian and Macedonian Slavic and East European Journal20273ndash279

Embick David 1995 lsquolsquoMobile inflectionsrsquorsquo in Polish In NELS 25 127ndash142 GLSA Universityof Massachu-setts Amherst

Embick David 1997 Voice and the interfaces of syntax Doctoral dissertation University of PennsylvaniaPhiladelphia

Embick David 2000 Features syntax and categories in the Latin perfect Linguistic Inquiry 31185ndash230Embick David and Roumyana Izvorski 1996 Participle-auxiliary word-orders in Slavic In Formal Ap-

proaches to Slavic LinguisticsThe Cornell Meeting 210ndash239 Michigan Slavic PublicationsUniver-sity of Michigan Ann Arbor

Emonds Joseph 1978 The verbal complex V cent -V in French Linguistic Inquiry 9151ndash175Emonds Joseph 1987 The Invisible Category Principle Linguistic Inquiry 18613ndash632Endzelotilde ns Janis 1971 Comparative phonology and morphology of the Baltic languages Trans by William

R Schmalstieg and Benjamin sIuml Jegers The Hague MoutonErnout Alfred and Francois Thomas 1951 Syntaxe latine Paris KlincksieckFlagg Elissa 2000 Aspect and the syntax of English imperatives Generals paper MIT Cambridge MassFranks Steven 2001 The internal structure of Slavic noun phrases with special reference to Bulgarian

Ms Indiana University BloomingtonFranks Steven and Tracy Holloway King 2000 A handbook of Slavic clitics Oxford Oxford University

PressGeniusIuml ieneCcedil Emma 1987 The typology of reflexives Berlin Mouton de GruyterHalle Morris 1990 An approach to morphologyIn NELS 20 150ndash184 GLSA University of Massachusetts

AmherstHalle Morris 1997 DistributedMorphologyImpoverishmentand fission In MIT working papers in linguis-

tics 30 PFmdashPapers at the interface425ndash449 MITWPL Department of Linguisticsand PhilosophyMIT Cambridge Mass

Halle Morris and Alec Marantz 1993 Distributed Morphology and the pieces of inflection In The viewfrom Building 20 Essays in linguistics in honor of Sylvain Bromberger ed Kenneth Hale and SamuelJay Keyser 111ndash176 Cambridge Mass MIT Press

Halpern Aaron 1992a The Balkan definite article and pseudo-second position In Proceedings of theEighteenth Annual Meeting of the Berkeley Linguistics Society 338ndash349 Berkeley Linguistics Soci-ety University of California Berkeley

Halpern Aaron 1992b Topics in the placement and morphology of clitics Doctoral dissertation StanfordUniversity Stanford Calif


Harley Heidi and Rolf Noyer 1998 Licensing in the non-Lexicalist lexicon Nominalizations Vocabularyitems and the encyclopaedia In MIT working papers in linguistics 32 Proceedings of the PennMIT Roundtable on Argument Structure and Aspect 119ndash138 MITWPL Department of Linguisticsand Philosophy MIT Cambridge Mass

Harley Heidi and Rolf Noyer 1999 Distributed Morphology Glot International 43ndash9Harris James 1991 The exponence of gender in Spanish Linguistic Inquiry 2227ndash62Hayes Bruce 1990 PrecompiledphrasalphonologyIn The syntax-phonologyconnectioned Sharon Inkelas

and Draga Zec 85ndash108 Chicago University of Chicago PressKalman Bela 1965 Vogul chrestomathy Bloomington Indiana University PressKester Ellen-Petra 1992 Adjectival inflection and dummy affixation in Germanic and Romance languages

In Papers from the Workshop on the Scandinavian Noun Phrase University of Umea Klavans Judith 1985 The independence of syntax and phonology in cliticization Language 6195ndash120Klavans Judith 1995 On clitics and cliticization New York GarlandKroch Anthony 1989 Function and grammar in the history of English Periphrastic do In Language change

and variation ed Ralph W Fasold and Deborah Schiffrin 133ndash172 Philadelphia John BenjaminsLaka Itziar 1990 Negation in syntax On the nature of functional categories and projections Doctoral

dissertation MIT Cambridge MassLeumann Manu Johann Baptist Hofmann and Anton Szantyr 1963 Lateinische Grammatik auf der Grund-

lage des Werkes von Friedrich Stolz und Joseph Hermann Schmalz 1 Band Lateinische Laut- undFormenlehre Munich Beckrsquosche Verlagsbuchhandlung

Marantz Alec 1984 On the nature of grammatical relations Cambridge Mass MIT PressMarantz Alec 1988 Clitics MorphologicalMerger and the mapping to phonologicalstructure In Theoreti-

cal morphology ed Michael Hammond and Michael Noonan 253ndash270 San Diego Calif AcademicPress

Marantz Alec 1992 Case and licensing In ESCOL rsquo91 234ndash253 CLC Publications Cornell UniversityIthaca NY

Marantz Alec 1995 A late note on Late Insertion In Explorations in generative grammar ed Young-SunKim 357ndash368 Seoul Hankuk Publishing Co

Marantz Alec 2000 Reconstruction of the lexical domain with a single generative engine Handout of atalk presented at MIT Cambridge Mass

Minkoff Seth 1994 Plurality clitics and Morphological Merger In MIT working papers in linguistics 20Papers from the 5th Student Conference in Linguistics 177ndash192 MITWPL Department of Linguis-tics and Philosophy MIT Cambridge Mass

Nevis Joel A and Brian D Joseph 1992 Wackernagel affixes Evidence from Balto-Slavic Yearbook ofMorphology 393ndash111

Noyer Rolf 1992 Features positionsand affixes in autonomousmorphologicalstructureDoctoral disserta-tion MIT Cambridge Mass

Noyer Rolf 1997 Features positions and affixes in autonomous morphological structure New YorkGarland

Noyer Rolf To appear Clitic sequencesin Nunggubuyuand PF convergenceNatural Languageamp LinguisticTheory

Perlmutter David 1971 Deep and Surface Structure constraints in syntax New York Holt Rinehart andWinston

Pesetsky David 1995 Zero syntax Experiencers and cascades Cambridge Mass MIT PressPollock Jean-Yves 1989 Verb movement Universal Grammar and the structure of IP Linguistic Inquiry

20365ndash424Sadock Jerrold 1991 Autolexical Syntax A theory of parallel grammatical representations Chicago

University of Chicago Press


Sandstrom Gorel and Anders Holmberg 1994 Adjective incorporation and the syntax of the Scandinaviannoun phrase In Reports from the Department of General Linguistics University of Umea 35 81ndash97University of Umea

Santelmann Lynn 1992 Den-support An analysis of double determiners in Swedish In Papers from theWorkshop on the Scandinavian Noun Phrase 100ndash118 University of Umea

Scatton Ernest 1980 The shape of the Bulgarian definite article In Morphosyntax in Slavic ed CatherineChvany and Richard Brecht 204ndash211 Columbus Ohio Slavica

Schoorlemmer Martin 1998 Possessors articles and definiteness In Possessors predicates and movementin the Determiner Phrase ed Artemis Alexiadou and Chris Wilder 55ndash86 Amsterdam John Benja-mins

Schutze Carson 1994 Serbo-Croatian second position clitic placement and the phonology-syntaxinterfaceIn MIT working papers in linguistics 21 Papers on phonology and morphology 373ndash473 MITWPLDepartment of Linguistics and Philosophy MIT Cambridge Mass

Senn Alfred 1966 Handbuch der litauischen Sprache Heidelberg Carl Winters UniversitatsverlagSommer Ferdinand 1914 Handbuch der lateinischen Laut- und Formenlehre Heidelberg Carl Winters

UniversitatsbuchhandlungSproat Richard 1985 On deriving the lexicon Doctoral dissertation MIT Cambridge MassStairs Emily F and Barbara E Hollenbach 1981 Gramatica huave In Diccionario huave de San Mateo

del Mar ed G A S Kreger and Emily F Stairs 283ndash391 Mexico City SILStang Christian S 1966 Vergleichende Grammatik der baltischen Sprache Oslo UniversitetsforlagetTaraldsen Knut Tarald 1990 D-projections and N-projections in Norwegian In Grammar in progress

GLOW essays for Henk van Riemsdijk ed Joan Mascaro and Marina Nespor 419ndash431 DordrechtForis

Tomic Olga MisIuml eska 1996 The Balkan Slavic nominal clitics In Approaching second Second positionclitics and related phenomena ed Aaron Halpern and Arnold Zwicky 511ndash536 Stanford CalifCSLI Publications

ZinkevicIuml ius Zigmas 1957 LietuviuIcirc kalbos iIcircvardzIumliuotiniuIcirc budvardzIumliuIcirc bruozIumlai Vilnius ValstybineCcedil politineCcedil sir mokslineCcedil s literaturos leidykla

Zwicky Arnold and Geoffrey K Pullum 1983 Cliticization vs inflection English nrsquot Language 59502ndash513

Department of Linguistics619 Williams HallUniversity of PennsylvaniaPhiladelphia Pennsylvania 19104ndash6305

embicklingupennedurnoyerlingupennedu


follow from syntactic principles (see Perlmutter 1971 also see Bonet 1991 and Noyer to appearfor explicit discussion in the framework assumed here) In the second domain which we explorehere it has also been recognized that syntactic movement cannot be responsible for certain move-ment operations One of the most familiar comes from the inflectional system of English Whileit has long been recognized that English main verbs do not move to T(ense) in the syntacticderivation (Emonds 1978 Pollock 1989) it is nevertheless the case that tense morphology appearson the verb in the surface string John kick-ed the ball to Mary The positioning of tense morphol-ogy on the verb has been attributed to Lowering a movement operation that ultimately derivesfrom the Affix Hopping transformation of early generative grammar Subsequent attempts toidentify and analyze lsquolsquomismatchesrsquorsquo between morphology and syntax can be found in the workof Marantz (1984 1988) and Sadock (1991) Our goal in this article is to develop a more compre-hensive theory of the types of morphological movement operations in the grammar and to situatethese operations within an explicit architecture for the PF derivation The focus is on cases inwhich some sort of displacement must apply to generate the proper phonological form from anidentifiable syntactic structure We thus exclude from consideration cases in which for examplethe syntactic distribution of clitic clusters is exactly what one would expect given the propersyntactic analysis1

1 Syntactic versus Postsyntactic

Our initial discussion of movement in PF has two components The first concerns why postsyntac-tic displacements are needed in addition to syntactic movement The second addresses the ideathat syntactic movements of the more familiar sort are part of the PF derivation as well

11 Clitic Placement

The need for nonsyntactic movements has been clearest in the domain of clitic placementmdashinparticular in the analysis of second-position clitics In some cases the question that was posedwas whether or not syntactic movement could possibly account for patterns of clitic placementor in particular second-position phenomena We take this question to be ill formed Syntax is agenerative system and assuming that one is willing to loosen many of the constraints on syntacticmovement the ability of such a system to capture certain linear orders should never have beenin question Rather the question that we take to be central is whether or not it is desirable tohave syntax perform such operations The two positions that we wish to contrast along with themeans by which each approach captures apparent postsyntactic movements are as follows

(1) a Syntax only Syntax performs operations that are explicitly executed so as to resolvea morphophonological problem Patterns of apparent postsyntactic movement arereducible to the effects of these lsquolsquospecialrsquorsquo syntactic processes

b PF movement Syntax generates and moves terminals according to its own principles

1 A third type of difference in structure between PF and syntax proper involves the addition of morphemes at PFto meet language-particular requirements this will be outlined below




12 Syntax in PF

















XP

X WP

Y

Z Y

X

gd

ab

1






















31 Lowering



[X P X0 [Y P Y0 ]] [X P [Y P [Y 0 Y0`X0 ] ]]












(9) [X P X [Y P [Z P Z] Y]]


(10) [X [Z Y]]




(11) [[Z 0 Z`X] Y]






(13) Z | [X P X Y W] Z | [X P Y`X W]


















4 A Model







PFLF BRANCHING

PHONOLOGICALFORM

macr

Lowering

LocalDislocation

(ProsodicInversion)




MORPHOLOGY

VocabularyInsertion





(17) [A B C] [B C`A] or [B A`C]






42 Late Lowering






she t2 t1


43 Summary






51 Lowering of Def













(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References







































































12 Syntax in PF

















XP

X WP

Y

Z Y

X

gd

ab

1






















31 Lowering



[X P X0 [Y P Y0 ]] [X P [Y P [Y 0 Y0`X0 ] ]]












(9) [X P X [Y P [Z P Z] Y]]


(10) [X [Z Y]]




(11) [[Z 0 Z`X] Y]






(13) Z | [X P X Y W] Z | [X P Y`X W]


















4 A Model







PFLF BRANCHING

PHONOLOGICALFORM

macr

Lowering

LocalDislocation

(ProsodicInversion)




MORPHOLOGY

VocabularyInsertion





(17) [A B C] [B C`A] or [B A`C]






42 Late Lowering






she t2 t1


43 Summary






51 Lowering of Def













(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References















































































XP

X WP

Y

Z Y

X

gd

ab

1






















31 Lowering



[X P X0 [Y P Y0 ]] [X P [Y P [Y 0 Y0`X0 ] ]]












(9) [X P X [Y P [Z P Z] Y]]


(10) [X [Z Y]]




(11) [[Z 0 Z`X] Y]






(13) Z | [X P X Y W] Z | [X P Y`X W]


















4 A Model







PFLF BRANCHING

PHONOLOGICALFORM

macr

Lowering

LocalDislocation

(ProsodicInversion)




MORPHOLOGY

VocabularyInsertion





(17) [A B C] [B C`A] or [B A`C]






42 Late Lowering






she t2 t1


43 Summary






51 Lowering of Def













(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References


















































































31 Lowering



[X P X0 [Y P Y0 ]] [X P [Y P [Y 0 Y0`X0 ] ]]












(9) [X P X [Y P [Z P Z] Y]]


(10) [X [Z Y]]




(11) [[Z 0 Z`X] Y]






(13) Z | [X P X Y W] Z | [X P Y`X W]


















4 A Model







PFLF BRANCHING

PHONOLOGICALFORM

macr

Lowering

LocalDislocation

(ProsodicInversion)




MORPHOLOGY

VocabularyInsertion





(17) [A B C] [B C`A] or [B A`C]






42 Late Lowering






she t2 t1


43 Summary






51 Lowering of Def













(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References












































































(9) [X P X [Y P [Z P Z] Y]]


(10) [X [Z Y]]




(11) [[Z 0 Z`X] Y]






(13) Z | [X P X Y W] Z | [X P Y`X W]


















4 A Model







PFLF BRANCHING

PHONOLOGICALFORM

macr

Lowering

LocalDislocation

(ProsodicInversion)




MORPHOLOGY

VocabularyInsertion





(17) [A B C] [B C`A] or [B A`C]






42 Late Lowering






she t2 t1


43 Summary






51 Lowering of Def













(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References






































































(11) [[Z 0 Z`X] Y]






(13) Z | [X P X Y W] Z | [X P Y`X W]


















4 A Model







PFLF BRANCHING

PHONOLOGICALFORM

macr

Lowering

LocalDislocation

(ProsodicInversion)




MORPHOLOGY

VocabularyInsertion





(17) [A B C] [B C`A] or [B A`C]






42 Late Lowering






she t2 t1


43 Summary






51 Lowering of Def













(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References



















































































4 A Model







PFLF BRANCHING

PHONOLOGICALFORM

macr

Lowering

LocalDislocation

(ProsodicInversion)




MORPHOLOGY

VocabularyInsertion





(17) [A B C] [B C`A] or [B A`C]






42 Late Lowering






she t2 t1


43 Summary






51 Lowering of Def













(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References






































































PFLF BRANCHING

PHONOLOGICALFORM

macr

Lowering

LocalDislocation

(ProsodicInversion)




MORPHOLOGY

VocabularyInsertion





(17) [A B C] [B C`A] or [B A`C]






42 Late Lowering






she t2 t1


43 Summary






51 Lowering of Def













(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References






































































42 Late Lowering






she t2 t1


43 Summary






51 Lowering of Def













(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References







































































51 Lowering of Def













(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References





































































(22) AdjNoun

DP

D

t

A

xubava

A

D N

NP

AP

ta kniga












lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References








































































lsquoyour bookrsquo






























(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References























































































(32) Clitic and Def

DP

D AP

D

ta

A NPCl

vi xubava N

kniga

DP

t AP

A NP

xubava D

D

ta

Cl

N

kniga

vi







53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References







































































53 Conclusions









61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References






































































61 Definitions





XP

X WP

Y

Z Y

X

hu

gd

ab



















(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References















































































(38) [X [[W W`Y] Z]]



















1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References


















































































1-TH-cut-REFL-PL







































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References








































































































(56) [-si [V T]] [[V Aring si T]]











(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References







































































(57) [V T] [[V 0 V Aring T]]








lsquothe mousersquo
























DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References






















































































DP

D

D AP

[def] A NP






(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References





































































(64) Processes

DP

D

D AP

[def]

[def]

d- A NP
















































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References













































































































T

T NegP

Neg

[neg] vP

eat the apples


T

T

Tv

NegP

Neg

[neg] vP

agree with Bill






(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References








































































(79) [X P (ZP) [X X (YP)]]









T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References








































































T

vPT

AdvP

v

v

Iuml P

v

completely






T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References






































































T

vPT

DP

[neg]

v

v

v

v

John

Iuml P


not









(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References





































































(84) dis-agree

v

v

X


Iuml P

Iuml P



8 Conclusions









References







































































References




































































movement operations after syntax - linguisticsembick/move.pdfsyntax david embick rolfnoyer we...

Documents